Submitted manuscript, 1.65 MB, PDF document
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Chapter (peer-reviewed) › peer-review
Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Chapter (peer-reviewed) › peer-review
}
TY - CHAP
T1 - Detecting jumps in high-frequency prices under stochastic volatility
T2 - a data-driven approach
AU - Tsai, Ping-Chen
AU - Shackleton, Mark
PY - 2016/5
Y1 - 2016/5
N2 - Detecting jumps in asset prices over a daily interval consists of testing for the significance of the difference between quadratic variation and integrated variance. Detecting jumps in high-frequency prices requires the additional tasks of estimating spot volatility and controlling for over-rejection due to multiple comparisons. We generalize two intraday tests commonly used in the literature and identify the test statistic that has the highest power at a given test level. The daily maximums of such test statistics admit an asymptotic generalized extreme value (GEV) distribution with a strictly positive shape parameter, as opposed to the limiting Gumbel distribution with a shape parameter zero for i.i.d. Gaussian maximums. The shape parameter of GEV distribution can thus be seen as a measure of bias correction for the test under stochastic volatility. We calibrate the shape parameter with a credible volatility model estimated from our data, which are Spyder (SPY) returns during January, 2002 and April, 2010. Empirical results are broadly consistent with those from simulation.
AB - Detecting jumps in asset prices over a daily interval consists of testing for the significance of the difference between quadratic variation and integrated variance. Detecting jumps in high-frequency prices requires the additional tasks of estimating spot volatility and controlling for over-rejection due to multiple comparisons. We generalize two intraday tests commonly used in the literature and identify the test statistic that has the highest power at a given test level. The daily maximums of such test statistics admit an asymptotic generalized extreme value (GEV) distribution with a strictly positive shape parameter, as opposed to the limiting Gumbel distribution with a shape parameter zero for i.i.d. Gaussian maximums. The shape parameter of GEV distribution can thus be seen as a measure of bias correction for the test under stochastic volatility. We calibrate the shape parameter with a credible volatility model estimated from our data, which are Spyder (SPY) returns during January, 2002 and April, 2010. Empirical results are broadly consistent with those from simulation.
KW - Jumps
KW - High-frequency Data
KW - Bi-power Variation
KW - Realized Variance
KW - Stochastic Volatility
KW - Multiple Tests Correction
KW - Extreme Value Theorem
KW - HAR Regression
UR - http://www.wiley-vch.de/publish/dt/books/forthcomingTitles/EC00/1-118-44398-5/?sID=4deb2lnm48aq6qvcsbuv34epp1
M3 - Chapter (peer-reviewed)
SN - 9781118443989
SP - 137
EP - 165
BT - Handbook of high-frequency trading and modeling in finance
A2 - Florescu, Ionut
A2 - Mariani, Maria C.
A2 - Stanley, H. Eugene
A2 - Viens, Frederi G.
PB - John Wiley
CY - Chichester
ER -